Twin Delayed Deep Deterministic Policy Gradient Algorithm for a Heterogeneous Multifactory Remanufacturing Optimization Problem

To reduce resource consumption and environmental impact, the manufacturing industry increasingly leans towards repurposing, repairing, or updating products. In a multifactory environment, considering the disassembly line balancing problem helps enterprises improve production efficiency and reduce costs. Thus, this work proposes a heterogeneous multifactory remanufacturing optimization problem, considering the disassembly techniques and U-shaped disassembly lines that are used in heterogeneous disassembly factories. A mixed integer programming model for profit maximization is established. Reinforcement learning methods open new avenues for addressing complex scheduling issues in actual production. This article utilizes the twin delayed deterministic policy gradient algorithm to solve the proposed problem. It validates the effectiveness of the algorithm by comparing it with CPLEX. Through various experimental cases, it demonstrates that this method achieves better convergence and higher profits compared to deep deterministic policy gradient, soft actor-critic, and advantage actor-critic algorithms.